Noise robust rotation invariant features for texture classification
Pattern Recognition
| Hi-index | 0.00 |
This paper performs a new method for analyzing regular and near-regular texture, which combines detection of repeated texels with extraction of spatial organization among these texels. For obtaining size, position and class of individual texel in a given texture, we apply an affine transform based method to estimate the similarity measurement between texels. In addition, Delaunay triangulation-like method is used to extract the triangular grid for describing the underlying texture periodicity. Our method can give further characterization for such textures when some useful information are computed, including the distance, direction of two adjacent texels. And the topology among all texels is also constructed for extracting the spatial neighborhood relationship of texels. We test our method on various sample textures, and give an extension on real images containing a type of such texture. Experimental results demonstrate that our method for regular and near-regular texture characterization is feasible and effective.